CoreML
Fewer than 5 people recommend it so far, so there's no written verdict. Here is what commenters said.
Why people recommend it
“It is very suitable and very straightforward for a crazy large amount of models!”
“I use CoreML for models designed for other platforms. I port the models to it but it works for that without much trouble.”
Show 2 more
“I would most likely bet on CoreML since it have support for ANE”
“better use inference on ANE (Apple Neural Engine) via CoreML rather than Metal”
Caveats and dissent
“CoreML treats it as a black-box scheduler, stripping away any direct control or ability to train”
“ANE and CoreML is kind of meh for most use cases”
Show 2 more
“Also generally I think CoreML isn't the best.”
“has had a few issues running models with CoreML”
Often compared with
Threads where it came up
- ONNX Runtime and CoreML May Silently Convert Your Model to FP16 2 mentions
- We turned EmbeddingGemma 2 into a speech-to-text model 1 mention
- Benchmarking Pocket-Scale Inference 1 mention
- Show HN: I trained a 125M model to autocomplete piano on-device 1 mention
- Apple Core AI Framework 1 mention
- Google Gemma 4 Runs Natively on iPhone with Full Offline AI Inference 1 mention
Also written as Core ML. Data: coreml.json. Quotes are verbatim; the verdict and theme labels are written by a model from these comments, and every clause links to its sources.
All comments 15
“It is very suitable and very straightforward for a crazy large amount of models!”
“traditional transfer learning like Apple's Core.ML”
“I’m using Core ML”
“I use CoreML for models designed for other platforms. I port the models to it but it works for that without much trouble.”
“requires using a specific pipeline specifically for CoreML”
“If anyone on CoreML has a principled explanation”
“I would most likely bet on CoreML since it have support for ANE”
“Rebuilt as an iOS/macOS app with CoreML”
“CoreML treats it as a black-box scheduler, stripping away any direct control or ability to train”
“ANE and CoreML is kind of meh for most use cases”
“the ANE toolchain from CoreML down through Espresso”
“better use inference on ANE (Apple Neural Engine) via CoreML rather than Metal”
“Also generally I think CoreML isn't the best.”
“has had a few issues running models with CoreML”
“when using the CoreML execution provider”